We have started by utilizing data generated at the Trans-NIH Center for Human Immunology (CHI) to assess the immune phenotypes of healthy individuals at baseline and after perturbations, particularly with seasonal influenza vaccination. The CHI has generated multiple types of measurements of peripheral blood mononuclear cells (PBMC), including microarray data for measuring transcript abundance, multiple panels of 15-color flow cytometry for assessing cell populations (and abundance of key markers), luminex assays for measuring serum cytokine concentrations, genome-wide genotyping, and immunological endpoints such as virus-specific antibody titers and B cell Elispots. We have successfully resolved a number of data analysis and modeling challenges and have been conducting integrative modeling projects using both in-house and public data sets to draw novel insights into human immunology. In addition to utilizing and integrating CHI and public data sets, we have initiated collaborative projects with both extramural and intramural colleagues by applying our human systems immunology approach to meta-analyze data from multiple cohorts as well as generating and integrating new data from both healthy and disease subjects. Recent highlights of our efforts include: 1. By utilizing vaccine perturbation data, we have developed a conceptual and methodological framework to quantify baseline and response variations at the level of genes, pathways, and cell populations in a cohort of individuals. Our framework takes advantage of such natural variations to systematically infer correlates, build predictive models of response quality after immune perturbation, and infer novel functional connections among various components in the human immune system. We have applied this framework to the influenza vaccination study utilizing antibody titer response as an exemplar endpoint. We confirmed previously known post-vaccination correlates based on gene expression and plasmablast frequency from day 7 samples. More importantly, using an approach that compensates for the influence of pre-existing serology, age and gender, we derived accurate predictive models of antibody responses using pre-vaccination data alone. This finding has obvious implications for the design of future vaccine trials and for developing a deeper understanding of the molecular and cellular parameters that contribute to robust vaccine responses. The robustness and translational potential of these findings is further emphasized by our discovery that the parameters playing the greatest role in correct response prediction are those with the most stable baseline values across individuals. This raises the prospect of monitoring immune health and predicting the quality of responses in the clinic via the evaluation of these baseline blood biomarkers. The conceptual and computational analysis framework we have developed can also be applied to systems and population level exploration in a number of medically relevant circumstances, including but not limited to the effects of drug intervention or natural disease history studies in humans. 2. By utilizing natural variation among human subjects, we have developed a de novo approach of inferring predictive models of cell population frequencies using gene expression data alone (e.g., those from heterogeneous tissues such as PBMCs and whole blood). Our approach infers such predictive models from gene expression and flow phenotyping data alone without a priori information on the gene expression pattern of individual cell types or subsets. By using baseline data from our vaccine study, we have successfully derived cell-frequency predictive models for more than 20 immune cell subsets spanning the T, B and myeloid lineages. We have applied these models to predict the cell frequency changes across more than 100 whole blood or PBMC gene expression comparisons involving diverse diseases with or without apparent immunological conditions. 3. Since we routinely use and analyze publicly available data to argument data generated in-house, we have developed a web-based framework (including both user interfaces and database components) to facilitate search, retrieval, annotating, meta-analysis, and gene-expression signature generation. At the core of our tool are interfaces for creating, annotating, and sharing (among the user community) of which groups of samples can be compared to form classical gene-expression signatures or expression difference profiles the latter can be used to integrate across data sets and studies to generate virtual perturbation profiles across all genes. Since annotating such comparison groups is often one of the most time-consuming steps of reusing existing data, our framework provides functions for users to share their own annotations and search for others annotations. Our tool was designed for experimental biologist to take full advantage of reusing and sharing large-scale data for obtaining biological insights. 4. Together with colleagues at the Human Immunology Project Consortium (HIPC), we have been performing meta-analysis of multiple human influenza vaccination data sets to derive common predictive signatures of vaccine responses using pre-vaccination gene-expression data. We have successfully uncovered both gene- and gene module-based predictive signatures for younger subjects using data from several study cohorts spanning multiple seasons and geographic locations.